We need to make sure we're not using so many cores that the tests
are starved of memory.

Signed-off-by: Adam Farley <[email protected]>

Instead of what it does right now, which is to compare bytes with
kilobytes, and incorrectly assume that the kilobytes number is always
smaller.

Signed-off-by: Adam Farley <[email protected]>

In cgroup v1 we used memory.limit_in_bytes to store the maximum
memory allocated to the container. In v2, we use memory.max. This
change allows us to check for both, includes a meminfo check for
non-containers, and adds a few debug comments so we can be sure this
new code is working.

My plan is to run tests on a diverse set of machines, and to remove
the debug statements prior to merging.

Signed-off-by: Adam Farley <[email protected]>

To ensure we're looking in the correct file for the maximum memory
size. Also to handle permissions issues, empty files, etc.

Signed-off-by: Adam Farley <[email protected]>

Signed-off-by: Adam Farley <[email protected]>

Makes it easier to maintain, and more resilient to failure.

I've also added a commented, formatted copy of the script to help
people read it in the future.

Signed-off-by: Adam Farley <[email protected]>

Signed-off-by: Adam Farley <[email protected]>

Signed-off-by: Adam Farley <[email protected]>